If you've ever tried to budget by hand, you know the real chore isn't the math — it's the sorting. Was that charge groceries or household? Is the coffee shop "dining" or "coffee"? Which of the 30 transactions this week were subscriptions? Manually tagging every purchase is tedious, and tedium is exactly what kills budgets.
Automatic categorization exists to remove that chore entirely. Plenty of apps describe this as "AI" and leave it there, which tells you nothing about what happens to your bank data. So let's open it up: here is exactly how TrackE5 sorts your spending, and why we deliberately chose not to use AI to do it.
The problem it solves
A bank transaction is surprisingly unhelpful on its own. It usually shows up as a cryptic merchant code, an amount, and a date. "SQ *MAPLE CAFE 604" doesn't tell a spreadsheet anything useful. A human can guess it's a coffee shop. Something has to make that guess for you, thousands of times, without you thinking about it.
How TrackE5 actually decides
Two things do the work, and neither of them is a model.
First, your bank connection provider — Plaid or Flinks — already resolves that messy merchant string into a recognisable merchant and a broad category. That resolution happens as part of the bank feed itself.
Second, TrackE5 maps that result onto our own category set using rules we wrote and maintain: mapping tables plus a rule engine that handles the cases where the provider's category is too broad or plainly wrong for how people actually budget.
The result is deterministic. The same merchant, with the same details, always lands in the same category. There is no probability, no temperature, no model weights — which means when a category looks wrong, it's a rule we can find and fix, not a black box we can only retrain and hope.
Why we don't use AI for this
We could route your transactions through a large language model. Some budgeting apps do exactly that, and their privacy policies name the AI vendor they send your financial data to. We decided against it, for three reasons.
- Your transactions stay with us. Sending spending data to a third-party inference provider means one more company holding a readable record of where you shop, what you earn and what you owe. We would rather not create that copy at all.
- Predictable beats clever. A budget is only useful if the numbers hold still. A model that recategorises the same coffee shop differently in March than it did in February quietly corrupts every trend you're looking at.
- We can explain every decision. If a charge lands in the wrong place, we can point at the rule responsible. That's a much better answer than "the model thought so."
How your corrections work
Everyone's spending is a little different — what's a "want" for one person is a recurring necessity for another. So when you correct a category, moving a charge from "dining" to "business," TrackE5 remembers that preference for that merchant, and applies it going forward.
That preference lives in your account only. It is never pooled with other customers' data, and it never trains anything that benefits somebody else. Your corrections tune your categories. That's all they do.
Why accurate categorization actually matters
This isn't just a convenience feature. The entire usefulness of a budget depends on the categories being right. If half your restaurant spending is mislabelled as groceries, your "food budget" is fiction and your decisions are based on bad data. Good categorization is what makes every downstream insight — your monthly totals, your category alerts, your spending trends — trustworthy.
Get the sorting right automatically, and budgeting transforms from a nightly data-entry task into something that simply happens in the background while you live your life.
How TrackE5 handles it
TrackE5 connects to your Canadian or American bank through Plaid and Flinks and imports your transactions automatically. As each one comes in, TrackE5 reads the merchant, amount and date, and matches it against a set of categorization rules — across CAD and USD natively. When it gets one wrong, you correct it once, and TrackE5 remembers your preference for that merchant. Prefer to add things manually or snap a receipt photo instead? That works too, and those are sorted the same way.
This runs on our servers, not on your device, and we think it's worth being specific about what that means. Your data is encrypted in transit and at rest and stored in Canada. The categorization rules are deterministic — no machine-learning model, and nothing sent to a third-party AI provider. Your transactions are never pooled with other customers' data and never used to train anything that benefits someone else.
The bottom line
Automatic categorization isn't magic, and it isn't a gimmick. It's a practical solution to the one thing that makes budgeting fail for most people: the effort of sorting. By reading the merchant, remembering your corrections, and handling the tedium automatically, it turns a clear picture of your finances from a weekend project into something you barely have to think about. And a budget you don't have to think about is a budget you'll actually keep.
